The Zhitong Finance App learned that Changjiang Securities released a research report stating that the optimization of humanoid robot hardware in the direction of cost reduction will not change the general trend of industry development. On the one hand, it indicates that the industry is gradually entering the mass production stage, so as to reduce mass production costs and promote better application penetration of humanoid robots, which is gradually receiving attention from manufacturers. On the other hand, it may indicate that current advances in operation and control algorithms have the ability to further optimize the cost of existing hardware architectures. The cost reduction trend in the industry will benefit domestic assembly and parts manufacturers. Domestic large-scale advantage is obvious. It has good cost reduction capabilities for parts that can be mass-produced, and can adapt to the global humanoid robot mass production release pace and cost reduction requirements.

Recommended attention: Screws, screw processing equipment, six-dimensional force sensors, harmonic speed reducers, etc. are expensive and require domestic alternatives to reduce costs.

The main views of Changjiang Securities are as follows:

Recently, the Nvidia GTC2024 conference came to an end. In his keynote speech, Huang Renxun showed the practical application of AI in the fields of robotics, warehousing, medical care, weather forecasting, etc., and was confident in the broad prospects for the future development of artificial intelligence. At the same time, FigureAI released the latest video. After connecting to the Open AI large model, Figure-1 showed excellent ability to use neural networks and multi-modal large models.

A number of robot manufacturers showed the actual application of humanoid robots, and the mass production process is getting closer

FigureAI recently released a video. After connecting to the big model of Open AI, Figure-1 can understand the implicit semantics of conversation, such as when a person expresses that they want to eat something, it can correspond to understanding giving apples other than other items. At the same time, it can continue to talk to people while picking up desktop trash according to instructions, showing the ability to operate in multiple threads. Second, Figure-1 moves very well and can actively adjust the motion mode according to the characteristics of different objects.

Open AI has expanded neural networks and multi-modal large models to robot perception, inference, and interaction. Figure-1 shows the application of excellent neural networks and multi-modal large models. Meanwhile, Tesla's recent video shows that Optimus has been able to walk steadily, turn smoothly, and fold clothes at the Tesla factory. Recently, Tesla continues to announce recruitment for AR/VR software engineers, data collection managers, technical managers, etc., and continues to promote mass production implementation. Domestic Preferred Choice (09880) Walker-S has begun practical training at a new energy vehicle factory.

Humanoid robot technology continues to be optimized. If it enters the mass production stage, we need to pay more attention to the cost reduction situation in the industrial chain

Currently, some manufacturers' humanoid robot structural designs include parts with high value and high processing and technical barriers. Typically, leading downstream manufacturers of planetary roller screws currently require high accuracy, difficult processing, and require high-end overseas machine tools for key process links. There are supply bottlenecks, and overall production costs are high. Downstream manufacturers are actively exploring cost reduction solutions. Previously, solutions included optimizing operation control algorithms to make up for shortcomings in hardware accuracy, etc., and using motor+planetary reducer structures to replace higher-value harmonic structures and linear actuator structures.

Currently, using brushless slotted motors as an alternative drive structure for dexterous hand hollow cup motors has become the direction of industry technology discussions. Hardware conversion will have trade-offs in terms of cost and performance. For example, when compared with brushless cored motors and hollow cup motors, brushless cored motors can obtain more torque with less power. The cost is relatively advantageous, but they also face certain limitations such as power loss, operating noise, and miniaturization. The hollow cup motor process is relatively complicated and the value is higher, but the motor efficiency is high (higher than traditional iron core motors) and the operation is stable and reliable. It has previously been used in the fields of medicine, robotics, etc., and is currently the main solution for smart humanoid robots.

Risk warning

The mass production implementation of humanoid robots fell short of expectations; the development of humanoid robot technology fell short of expectations; demand for humanoid robot products fell short of expectations.